I am assessing the extra social cost of caregiving by considering the labor changes and mental burden experienced by spouses and other family members who provided unpaid/informal care to patients. This evaluation is conducted based on the cost per quality-adjusted life years (QALY) of the caregivers, which replaced the common practice of using willingness-to-pay (WTP) per QALY for the patient as a common standard. My research initially focus on first-time acute stroke survivors with long-term care requirements.

In our evaluation of COPD exacerbation, we examined its characteristics and conducted race-specific assessments. Through logistic regression, we observed a significant difference in exacerbation rates among different races, even after controlling for medication intake and relevant behavior changes. Additionally, we analyzed spirometry features with a particular focus on immune standards using an ordered logit model to assess exacerbation levels.

By integrating the Nurse-initiated Protocol Order (NIPO) and implementing the concept of "music chairs," we successfully optimized the allocation of "runners" across various testing locations and improved the efficiency of lab test processing. This initiative effectively mitigated the bullwhip effect observed in surgical queues within the emergency department. Consequently, we achieved COVID-19 testing results for over 85% of patients in under 3 hours when golden standard was set at 24 hours.

Utilizing a two-stage model incorporating bowel gas as a factor, we applied a random forest model to predict the location of pancreatic cancer tumors within the detectable range using the breath-holding method. Our team achieved an accuracy rate of over 75% in these predictions.

Through the integration of results obtained from Propensity Score Matching and Difference in Differences Models in our previous project, we conducted a reassessment of hospital stay costs. Our findings indicate that for every $1 invested in the implementation of AI tools alongside process improvements over a period of 5 years, there is an estimated return of $5.3-6.4 derived from a comprehensive cost-benefit analysis, considering different model assumptions over three commonly applied transformational techniques.

Through implementing both Propensity Score Matching and Difference in Differences Models, our analysis revealed that Johns Hopkins University's emergency department is treating patients with acute ischemic stroke who are in a more critical condition, thereby explaining the comparable mortality rates observed. The results indicate a significant improvement of at least 0.21-0.25 Quality-Adjusted Life Years (QALYs) when comparing the modified Rankin Score (mRS) outcomes to the original outcomes.

Using random forest models, we made predictions on whether children would visit pediatric orthopedics for fractures based on precipitation and calendrical data. Our model achieved an accuracy of 79%. Upon obtaining the results, we adjusted the time block of schedules by shifting it 30 minutes earlier, resulting in a 31% decrease in length of stay (LOS). This reduction in LOS translates to a cost savings of $550 per patient per visit, as determined by running a median regression analysis.

Using Item Response Theory, I matched questionnaire responses with monitored bank account transactions for Vietnamese young adults. My findings indicate that children from different cities provided varying responses. After controlling for socioeconomic background, the initially claimed 11% improvement attributed to the specific training program (S2S) decreased significantly to 2-4%. However, a consistent growth trend in numbers has been evident since 2014.

By incorporating oil expenditure adjustments for endogenous effects and conducting interviews with over 200 part-time drivers, I built OLS models revealing a 25% crowd-out effect on "dispatching drivers." However, the impact on "running drivers" was found to be insignificant. Moreover, our findings suggest a minimum multiplier gain of 3.2 times per dollar loss experienced by taxi drivers in Taiwan by 2015-16.